smolm-autoreg-bpe-counterfactual-babylm-only_indef_articles_with_pl_nouns_removal-1e-3

This model was trained from scratch on the kanishka/counterfactual-babylm-only_indef_articles_with_pl_nouns_removal dataset. It achieves the following results on the evaluation set:

• Loss: 3.4114
• Accuracy: 0.4119

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.001
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 32000
• num_epochs: 20.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.5998	1.0	18600	3.7955	0.3595
3.3776	2.0	37200	3.5874	0.3805
3.245	3.0	55800	3.4956	0.3923
3.1698	4.0	74400	3.4301	0.3991
3.1095	5.0	93000	3.4080	0.4017
3.0618	6.0	111600	3.3783	0.4047
3.0262	7.0	130200	3.3656	0.4063
2.9992	8.0	148800	3.3350	0.4088
2.9653	9.0	167400	3.3531	0.4103
2.9376	10.0	186000	3.3526	0.4110
2.9136	11.0	204600	3.3538	0.4098
2.8922	12.0	223200	3.3425	0.4120
2.8698	13.0	241800	3.3346	0.4124
2.8466	14.0	260400	3.3660	0.4110
2.8253	15.0	279000	3.3566	0.4127
2.8058	16.0	297600	3.3781	0.4113
2.7908	17.0	316200	3.3851	0.4119
2.7701	18.0	334800	3.3872	0.4128
2.7511	19.0	353400	3.4038	0.4120
2.7292	20.0	372000	3.4114	0.4119

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1